🌈

Rainbow AI

Intelligent WhatsApp Assistant for Hospitality

User Manual — v2.0

Your complete guide to setting up, training, and managing your AI-powered WhatsApp concierge that handles guest inquiries 24/7.

📖 Table of Contents

1 What is Rainbow AI? 2 Dashboard Overview 3 Live Chat & WhatsApp 4 Training AI Understanding 5 Smart Routing 6 Responses & Knowledge Base 7 Chat Simulator & Testing 8 Performance & Analytics 9 Settings & Configuration 10 FAQ & Tips
🌈
Section 1

What is Rainbow AI?

Rainbow AI is an intelligent WhatsApp assistant designed specifically for the hospitality industry. It automatically handles guest messages on WhatsApp — answering questions about rooms, amenities, check-in procedures, pricing, WiFi access, and much more — in multiple languages including English, Malay, and Chinese.

Think of Rainbow AI as your 24/7 virtual front desk assistant that never sleeps, never takes a day off, and responds to every guest message within seconds. It understands what guests are asking, provides accurate answers from your knowledge base, and seamlessly escalates complex issues to your human staff when needed.

24/7
Always Available
3+
Languages Supported
<2s
Response Time
96%
Intent Accuracy
💬

Automatic Guest Replies

Instantly answers common questions about WiFi passwords, check-in times, room prices, facilities, and directions — without staff involvement.

🌐

Multi-Language Support

Communicates fluently in English, Bahasa Malaysia, and Chinese. Automatically detects the guest's language and responds accordingly.

🤖

Smart AI Classification

Uses a 4-tier intelligence system to understand guest messages with high accuracy. From keyword matching to advanced AI models — it gets smarter over time.

👥

Live Chat & Takeover

Staff can monitor all conversations in real-time. Switch to manual mode at any time to handle special requests personally.

Escalation System

Automatically routes emergencies, complaints, and complex requests to your staff. Never miss an urgent guest issue.

📊

Performance Analytics

Track response times, satisfaction rates, message volumes, and AI accuracy. Data-driven insights to continuously improve guest experience.

📈
Section 2

Dashboard Overview

The Dashboard is your home screen and control center. Every time you open Rainbow AI, this is the first thing you see. It gives you a quick overview of everything that matters: WhatsApp connection status, AI model health, recent messages, and quick shortcuts to common tasks.

Rainbow AI Dashboard
Figure 2.1 — Rainbow AI Dashboard showing WhatsApp status, AI models, quick stats, and server health

What You’ll See on the Dashboard

🚀

Quick Setup Checklist

When you first start, a guided checklist walks you through connecting WhatsApp, training your first intent, and testing the AI. You can dismiss it once done.

📱

WhatsApp Instances

Shows your connected WhatsApp number(s), their status (online/offline), and when they were last active. Add new numbers or manage existing ones here.

🤖

AI Models Status

Displays all configured AI models, their readiness status, and response times. A green indicator means the model is ready; you can click “Settings” to configure them.

📊

Quick Stats

Four key metrics at a glance: Messages Handled, Intent Accuracy, Average Response Time, and Guest Satisfaction Rate.

✅ Quick Actions

The Dashboard provides three shortcut buttons to your most common tasks:

  • Pair WhatsApp — Opens the QR code scanner to connect a new WhatsApp number
  • Train Intent — Jumps to the Understanding tab where you can teach the AI new skills
  • Test Chat — Opens the Chat Simulator so you can test how the AI responds

At the bottom, the Server Connection panel shows the health of all three system components (MCP Server, Backend API, and Frontend), while Recent Activity provides a live feed of the latest events.

💬
Section 3

Live Chat & WhatsApp

The Live Chat tab is where you manage all real-time WhatsApp conversations with your guests. It works just like WhatsApp Web, but with powerful AI features built in. You can monitor conversations, take over from the AI, and manage guest information — all from one screen.

Rainbow AI Live Chat
Figure 3.1 — Live Chat showing guest conversations with WhatsApp-style message bubbles, search, and filters

Three Response Modes

Rainbow AI gives you full control over how it handles conversations. Choose the mode that fits your needs:

Mode Icon How It Works Best For
Autopilot ✈️ AI reads the guest message and sends a reply automatically. No staff intervention needed. After-hours, routine questions (WiFi, directions, pricing)
Copilot 🤝 AI drafts a suggested reply, but waits for you to review and approve before sending. You can edit the suggestion. Important conversations, complex bookings, new staff training
Manual ✍️ AI stays silent. You write all replies yourself. A “Help me” button is available if you need AI assistance. Sensitive issues, complaints, VIP guests

Key Features

🔍

Search & Filters

Search conversations by guest name or phone number. Filter by date range, or use quick chips: All, Unread, Favourites, or Groups.

🌐

Translation Mode

Toggle the translation icon to automatically translate messages. Choose the target language and see both the original and translated text.

👤

Contact Details

Click a guest’s name to open their profile sidebar. View and edit their name, email, check-in/out dates, room assignment, and internal notes.

📎

File Sharing

Send photos, documents, and videos to guests directly from the chat. Perfect for sharing maps, receipts, or promotional materials.

💡 How to Switch Response Modes

In the chat header (top right area), you’ll see the current mode displayed next to the guest’s name. Click on it to switch between Autopilot, Copilot, and Manual modes. The change takes effect immediately for that conversation.

🧠
Section 4

Training AI Understanding

The Understanding tab is the brain of Rainbow AI. This is where you teach the AI to recognize what guests are asking about. Rainbow uses a powerful 4-tier classification system that combines speed with intelligence.

Understanding / Intent Detection
Figure 4.1 — Understanding tab showing the 4-tier intent detection system with test console and configuration templates

The 4-Tier Intelligence System

When a guest sends a message, Rainbow processes it through four layers of understanding, from fastest to most intelligent:

T1

🚨 Priority Keywords (Emergency Patterns)

Lightning-fast pattern matching for urgent situations like emergencies, fire, or medical issues. Processes in under 0.1 milliseconds.

Speed: ~0.1ms
T2

🔎 Smart Matching (Keywords)

Fuzzy keyword matching that catches common guest inquiries like “wifi password”, “check in time”, or “how much room”. Works across English, Malay, and Chinese keywords.

Speed: ~1ms
T3

📚 Learning Examples (Semantic Matching)

Uses AI embeddings to understand the meaning behind messages, even if the exact words don’t match your keywords. Train it with example phrases and it learns similar ones automatically.

Speed: ~50ms
T4

🤖 AI Fallback (Large Language Model)

For complex or ambiguous messages, a full AI model analyzes the conversation context and provides an intelligent response. This is the most powerful tier.

Speed: ~100–500ms

🌟 Quick Setup Templates

Don’t want to configure each tier manually? Choose from 7 pre-optimized templates that set up the entire system in one click:

  • T3 Balanced (Default — recommended) — 92–95% accuracy, great balance of speed and quality
  • T1 Maximum Quality — 95–98% accuracy, uses more AI resources
  • T2 High Performance — Fastest possible responses, keyword-only
  • T5 Tiered-Hybrid — Research-backed optimal settings
  • …and more for special cases like emergencies or multi-language focus

Test Console

The Test Console (always visible on the right side) lets you instantly test how Rainbow classifies any message. Type a phrase like “What’s the WiFi password?” and see exactly which tier caught it, what intent was detected, and with what confidence level. This is invaluable for verifying your training changes.

🔌
Section 5

Smart Routing

Smart Routing is the decision-making layer that determines what happens after Rainbow understands a guest’s message. While the Understanding tab figures out what the guest is asking, Smart Routing decides what action to take.

Smart Routing
Figure 5.1 — Smart Routing showing intent-to-action mapping with visual routing configuration

Routing Actions

For each recognized intent, you can assign one of these actions:

Action Description Example Use Case
Static Reply Sends an instant, pre-written message “What’s the WiFi password?” → sends WiFi info immediately
LLM Reply AI generates a personalized response using your knowledge base “Tell me about nearby restaurants” → AI crafts a helpful answer
Start Workflow Launches a multi-step interactive process “I want to book a room” → starts booking conversation
Escalate to Staff Forwards the conversation to a human operator “I have a complaint” → staff notification

💡 How It Works Together

Guest sends “I need to book a capsule for tonight” → Understanding detects the booking intent → Smart Routing maps this to “Start Booking Workflow” → Rainbow begins the multi-step booking conversation automatically.

💬
Section 6

Responses & Knowledge Base

The Responses tab is where you manage everything Rainbow AI says to guests. From quick one-line replies to complex multi-step workflows, this is your content control center.

Responses & Knowledge Base
Figure 6.1 — Responses tab with sub-tabs for Quick Replies, System Messages, Workflows, and Knowledge Base

Four Sub-Sections

Quick Replies

Pre-written answers for common questions. Organized by category (General, Pre-arrival, Check-in, During Stay, Check-out, Issues). Each reply supports multiple languages.

⚙️

System Messages

Technical messages like error notices, rate limit warnings, “thinking” placeholders, and default fallback responses. These ensure smooth communication even in edge cases.

🛠️

Smart Workflows

Multi-step interactive conversations for complex tasks like bookings, check-ins, and complaint handling. Visual workflow builder with conditions and actions.

📚

Knowledge Base

Your property’s information library. Edit markdown files covering pricing, facilities, WiFi, rules, location, and more. The AI uses these to generate informed answers.

📚 Knowledge Base — Your AI’s Memory

The Knowledge Base uses a progressive loading system that only loads relevant topic files when answering a guest’s question. This means the AI uses up to 60–70% fewer tokens compared to loading everything at once, making it both faster and more cost-effective. You can edit knowledge files directly in the browser with a live preview.

🧪
Section 7

Chat Simulator & Testing

The Chat Simulator is your testing playground. Before deploying any changes to production, you can test how Rainbow responds to different guest messages. It offers two modes: Quick Test for rapid single-message testing, and Live Simulation for monitoring actual WhatsApp conversations.

Chat Simulator
Figure 7.1 — Chat Simulator with Quick Test mode, Autotest suite, and session management

Quick Test Mode

Type any guest message and see how Rainbow would respond. The interface shows you:

Automated Testing Suite

Click the “Autotest” button to run a comprehensive test suite of 58 scenarios covering all six guest journey phases — from pre-arrival inquiries to post-checkout feedback. The test suite runs automatically and reports:

Passed

AI responded correctly with the expected intent and acceptable response quality.

⚠️

Warning

Response was acceptable but might need improvement (e.g., low confidence).

Failed

The AI gave an incorrect or unexpected response. Review and retrain needed.

👉 Best Practice: Test Before You Deploy

After making any changes to intents, keywords, responses, or routing, always run the Autotest suite to verify nothing was broken. You can also export test results as reports for record-keeping.

📊
Section 8

Performance & Analytics

The Performance tab gives you data-driven insights into how well Rainbow AI is serving your guests. Track satisfaction rates, response times, intent accuracy, and identify areas for improvement.

Performance Analytics
Figure 8.1 — Performance dashboard with feedback analytics, intent accuracy, and date range filtering

Key Metrics

👍

Guest Satisfaction

Tracks thumbs-up vs thumbs-down feedback from guests. See overall satisfaction rate and breakdown by intent category.

🎯

Intent Accuracy

Shows how accurately Rainbow is classifying guest messages. Identifies which intents are strong and which need more training.

Response Time

Average, P95, and P99 latency metrics. Know exactly how fast your AI is responding and spot any slowdowns.

📈

Message Volume

Hourly and daily message trends, peak hours analysis, and breakdown by intent type. Plan your staffing around real data.

💡 Feedback Collection

You can configure automatic feedback collection in the Performance tab settings. Set how often to ask guests for feedback, which intents to skip, and customize the feedback prompt in multiple languages. This data flows directly into your satisfaction analytics.

⚙️
Section 9

Settings & Configuration

The Settings tab is where you configure the technical backbone of Rainbow AI. It’s organized into three sub-sections for easy management.

Settings
Figure 9.1 — Settings showing AI model configuration, notifications, and operator management

Configuration Areas

🤖

AI Models

Configure your AI providers (Google Gemini, Groq, Ollama, etc.). Set API keys, choose default models, configure fallback chains, and test connections. Each model shows its status, speed, and readiness.

🔔

Notifications

Set up alerts for WhatsApp disconnections, server issues, and critical errors. Configure staff phone numbers and notification frequency.

👥

Operators

Manage your staff team. Add operators with their phone numbers and roles (Admin, Moderator, Viewer). Configure who receives escalation messages.

🛡️ AI Model Fallback Chain

Rainbow AI supports automatic failover between AI models. If your primary model (e.g., Google Gemini) is temporarily unavailable or slow, Rainbow automatically switches to the next model in your fallback chain. This ensures your guests always get a response, even if one provider has issues. You can configure the order of fallback models in the AI Models settings.

Section 10

Frequently Asked Questions

How do I connect my WhatsApp number?

Go to the Dashboard, click “+ Add Number”, enter your phone number and a label, then scan the QR code with your phone’s WhatsApp app (Linked Devices → Link a Device). The connection is established within seconds.

Can Rainbow AI handle multiple languages in the same conversation?

Yes! Rainbow automatically detects the language of each incoming message and responds in the same language. It supports English, Bahasa Malaysia, Chinese, and more. You can also manually translate messages using the translation toggle in Live Chat.

What happens when Rainbow doesn’t understand a message?

If the AI’s confidence is too low across all four tiers, it either sends a polite fallback message asking the guest to rephrase, or automatically escalates to a staff member. You can configure these thresholds in the Understanding tab.

How do I improve the AI’s accuracy?

Go to Understanding and add more keywords (T2) and training examples (T3) for the intents that need improvement. Use the Test Console to verify your changes, then run the Autotest suite in the Chat Simulator to confirm nothing else was affected.

Can I take over a conversation from the AI?

Absolutely. In Live Chat, simply switch the response mode from “Autopilot” to “Manual” or “Copilot”. You can take over at any point and the AI will stop responding for that conversation until you switch it back.

Is my data secure?

Rainbow AI runs on your own infrastructure. Conversations, guest data, and configuration stay on your servers. AI model API calls send only the necessary message context to generate responses — no bulk data sharing.

What if an AI model provider goes down?

Rainbow AI’s automatic fallback chain handles this seamlessly. If your primary model is unavailable, it automatically tries the next model in your configured chain. You can set up multiple fallback models in Settings → AI Models.

How do I add information about my property?

Go to Responses → Knowledge Base. Here you can edit markdown files covering topics like pricing, facilities, WiFi, house rules, nearby attractions, and more. The AI references these files when generating responses.

💡 Tips for Getting the Most Out of Rainbow AI

  1. Start with the T3 Balanced template It provides the best balance of accuracy and speed for most properties. You can fine-tune later.
  2. Add keywords in all three languages For each intent, add keywords in English, Malay, and Chinese. This dramatically improves detection for multilingual guests.
  3. Test after every change Use the Chat Simulator’s Autotest to verify your changes work correctly before they affect real guests.
  4. Review Performance weekly Check the Performance tab regularly to spot trends, identify weak intents, and track satisfaction improvements.
  5. Keep your Knowledge Base updated When prices change, new facilities open, or policies update, edit the Knowledge Base files so the AI gives accurate information.
  6. Use Copilot mode for training new staff New team members can see the AI’s suggested responses and learn the correct way to handle different guest inquiries.